NAVER’s upgraded search engine with ‘Hyperscale AI Language Model’ provides users with a new experience
NAVER’s upgraded search engine with ‘Hyperscale AI Language Model’ provides users with a new experience
NAVER’s upgraded search engine with ‘Hyperscale AI Language Model’ provides users with a new experience
- NAVER is expecting to reduce the rate of users encountering no results on the Search Engine Results Page (SERP) by 50 percent this year through the application of its ‘Hyperscale AI Language Model’
- The improved search feature, which analyzes users’ previous searches to provide keyword suggestions, will be implemented in Q3
- The new language model has also been further optimized to overcome processing speed degradation due to the vast amount of data
2021-05-11
NAVER’s Hyperscale Korean AI language model —the first of its kind in Korea—was applied to its search services on May 6.
The AI Language model offers a guided search experience, by correcting typos or providing keyword suggestions for more accurate search results. This serves as an essential feature to improve users’ search experience as it can reduce the chances of encountering no results on the Search Engine Result Page (SERP), preventing users from receiving the information they seek.
By applying the ‘Hyperscale Korean AI language model’ to the search engine, NAVER can propose alternative search queries in cases where the terms do not make sense. For example, if a user searches for “dandelionfloweerandwatercressdifferentfolwer,” NAVER will automatically correct it to “difference between a dandelion flower and watercress flower” for more precise search results.
“It was initially challenging to predict what search queries a user would enter into our search engine each time, so it was not easy to increase user satisfaction of the guided search feature above a certain level,” said Seonhoon Kim, Project Lead for Language Model Search Application at NAVER. “However, by applying the Hyperscale AI language model, we look forward to improving search query prediction and reducing the rate that our users encounter no results on the Search Engine Results Page (SERP) by 50 percent this year.”
Additionally, another search feature that predicts and recommends the next search query based on previous searches will also be introduced in the third quarter of 2021. For example, if a user searches for “Olive oil benefits” after searching for “How to make pasta,” NAVER will suggest the search query “How to make olive oil pasta.”
“In the past, hyperscale AI language models have received a lot of attention from academics for their potential to produce higher quality search results. But using the AI language model in search engine platform with a large user base requires the support of advanced technology such as model optimization and improved data processing speeds,” said In-ho Kang, Head of Natural Language Processing at NAVER. “Beginning with this keyword suggestion service, we will continue to strengthen our research and development in relevant areas to apply cutting-edge AI technology to our various services to offer users a better experience.”
NAVER will be hosting a technology conference at the end of May 2021. More details about the application of the AI language model will be shared during the conference.
-END-